See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/225697182

Analysis of stainless steel samples by energy dispersive X-ray fluorescence

(EDXRF) spectrometry

Article  in  Bulletin of Materials Science · December 2001

DOI: 10.1007/BF02704012

CITATIONS READS

31 1,576

3 authors:

M. K. Tiwari Ajit Kumar Singh

Raja Ramanna Centre for Advanced Technology Raja Ramanna Centre for Advanced Technology
139 PUBLICATIONS   1,264 CITATIONS    32 PUBLICATIONS   284 CITATIONS   

SEE PROFILE SEE PROFILE

Kawal Sawhney
Diamond Light Source
220 PUBLICATIONS   3,030 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Instrument and Technique Development View project

Interface Sharpening in Miscible and Isotopic Multilayers: Role of Short-Circuit Diffusion View project

All content following this page was uploaded by M. K. Tiwari on 28 May 2014.

The user has requested enhancement of the downloaded file.

Bull. Mater. Sci., Vol. 24, No. 6, December 2001, pp. 633–638. © Indian Academy of Sciences.

Analysis of stainless steel samples by energy dispersive X-ray

fluorescence (EDXRF) spectrometry
M K TIWARI*, A K SINGH and K J S SAWHNEY
Synchrotron Utilization Section, Centre for Advanced Technology, Indore 452 013, India

MS received 21 April 2001

Abstract. A simple method for the analysis of stainless steel samples is presented which is based on
radioisotope excited energy dispersive X-ray fluorescence (EDXRF) spectrometry and does not require any
type-standards. Both absorption and enhancement effects have been taken into account in the fundamental
parameter method for quantitative analysis and an iterative approach is followed for calculation of
concentrations in steel samples. Non-linear least square fitting (NL-LSF) procedures have been used to
determine accurately the fluorescent peak intensities. The method has been tested by analysing several CRM
standard reference samples and 304 and 316 steel samples assuming as unknown. The EDXRF results have also
been compared with the results of analysis of same samples by vacuum emission spark spectrometry (VES).
Obtained values for concentration in steel samples match quite well with their certified values.

Keywords. XRF; spectrometry; least squares fitting; stainless steel.

1. Introduction some what complex as not only the matrix absorption-

Energy dispersive X-ray fluorescence (EDXRF) technique
has become a powerful technique for non-destructive
multielement analysis of materials. Over the last years,
several developments have taken place in the methodology
for quantitative analysis using this technique with the aim,
on the one hand, to enhance the quality and reliability of
the results, and to extend the range of its applications on
the other hand. With the use of empirical and semi
empirical algorithms (Lachance and Traill 1966; Claisse
and Quintin 1967; Rasberry and Heinrich 1974), the
elemental composition of the sample is determined by
measuring the relative intensities of the X-ray fluorescent
peaks with that of a set of multielement standards or pure
elements. The comparative standards are desirable to be
similar in shape, form and concentration to the unknown
sample that sometimes limits the application of EDXRF
due to the non-availability of such standards. The funda-
mental parameter method (FP) using iterative approach is
a fast method (Sparks 1975; Sawhney and Lodha 1989)
which can work even when type-standards are not
available. This method gives reliable results for both trace
and major element analysis provided the absorption and
enhancement by the sample matrix is accounted for
properly (Van Dyck et al 1986). The analysis of multi-
component alloys like stainless steel (SS), in which both
trace and bulk elements are required to be determined, is
*Author for correspondence
enhancement effects have to be corrected, but accurate
peak intensities have to be determined from the complex
spectra using some fitting procedures. Generally in
stainless steel (SS) samples Ti(0⋅03–0⋅08), Si(0⋅03–1⋅1),
C(0⋅03–0⋅06), P(0⋅005–0⋅035), S(0⋅005–0⋅012), W(0⋅02–
0⋅3%) etc are present as minor components while
Cr(3⋅2–24⋅4), Ni(3⋅7–21⋅5), Fe(50⋅0–70⋅0), Cu(0⋅6–2⋅2)
and Mo(1⋅5–4⋅0%) are present as major components.
Barrea and Mainardi (1998) have analysed the stainless
steel samples by measuring two fluorescence spectra of
same sample under two different excitation conditions,
which provides a system of non-linear equations. Because
two measurements are required to be done for each
sample, the analysis time gets doubled in this method and
moreover this method cannot be used in radioisotope
excited XRF where the excitation energy is quasi-
monochromatic. They could determine elements Cr, Fe,
Ni, Nb, Mo only. They could not determine the concen-
tration of minor elements like Mn, Co, Cu etc whose Kα
peaks strongly overlapped with the neighbouring major
peaks.
In the present paper, we present a method for stainless
steel analysis which is simple, fast, non-destructive and
does not require any type-standards. We have determined
elements Cr, Mn, Fe, Co, Ni, Cu, Nb, Mo etc in the
stainless steel samples. The quantitative analysis has been
performed by our in-house developed CATXRF program
(Tiwari and Sawhney 2000), which is based on FP method
and corrects for both absorption and enhancement effects.
The peak areas have been determined by the use of non-
633
634 M K Tiwari, A K Singh and K J S Sawhney

linear least square fitting (NL-LSF) procedures. Only one energies E0, Ej and Ek respectively, Wk is the weight
single element standard has been used for the determina- fraction of the kth element, and M is mass of sample in
tion of system geometry factor. To verify the method, it g/cm2.
has been applied for the analysis of several certified
reference materials (CRM) and SS-304 and SS-316
3. Experimental
samples and the results obtained from our method have
also been compared with the analytical results obtained by
Several CRM stainless steel samples and unknown
the analysis of same samples by vacuum spark emission
samples were analysed on our EDXRF spectrometer
spectrometry (VES).
(Sawhney et al 1998) consisting of a Si(Li) detector
having an energy resolution of 155 eV at 5⋅9 keV. A
20 mCi Cd109 annular radioisotope was used as an
2. Theory
excitation source. The experimental setup is shown sche-
matically in figure 1. A 3 mm thick aluminum mask
For monochromatic excitation, the measured X-ray
having a centred hole of diameter 24⋅7 mm was used to
fluorescence intensity, Ij, is related to the analyte mass, mj,
ensure same illuminated surface area for all the samples.
by (Sparks 1975)
The sample-detector and excitation source-sample
(Ij)k = I0G ⋅ Kj ⋅ A ⋅ mj ⋅ (1 + Hjk), (1) distances were optimized to get maximum count rate
in the fluorescent peaks. Each sample was analysed for
where Kj is the relative efficiency to excite and detect the
1000 sec. The EDXRF spectra of SS samples are quite
fluorescent X-rays from element j and is given by
complex because of the presence of several small peaks
 1  from minor elements occurring near the big peaks of
Kj = τ 1 −  ⋅ω k ⋅ ε ⋅T ⋅ f , (1a) major elements and a simple deconvolution procedure
 J K , L  cannot be used to derive the peak areas of such small
peaks. Examples of such strong peak overlapping are the
A is the mass absorption correction factor and for
pairs of spectral lines Mn Kα–Cr Kβ, Co Kα–Fe Kβ and
infinitely thick samples, in which domain the stainless
Cu Kα–Ni Kβ . Therefore to derive accurate peak areas of
steel samples fall, is given by
such strongly overlapped peaks, we have used non-linear
1 least square fitting method using ORIGIN software
A= , (1b)
[ µ ( E0 ) cos ec(ψ 1 ) + µ ( E j ) cos ec(ψ 2 )] (Microcal Origin USA, 1995, Technical Graphics and
Data Analysis Software). Two typical fluorescence
Hjk is the enhancement factor and its value is zero for spectra of T-216 and SS-304 samples are shown in
elements that are not enhanced by any other sample figures 2 and 3. The system geometrical factor I0G was
constituent. Hjk for infinitely thick samples is given by determined by analysing a single element synthetic
Sparks (1975) as standard. To confirm the value of I0G, a few more
synthetic standards of different elements in various
1 m  1 
Hjk = ∑
2τ i ( E0 ) k =1
Wk ⋅ ω k ⋅ 1 −  ⋅ µ j ( E k ) ⋅ µ k ( E0 )

 J kl 

  µ (E0 )   µ (E j ) 
 log  1 +  log  1 + 

  µ ( E ) ⋅ sin ϕ 1   µ ( E k ) ⋅ sin ϕ 2  (1c)
• +
k
µ (E0 ) µ (E j ) ,
 
 sin ϕ 1 sin ϕ 2 

where j, k = index of the analyte and enhancing element;

E0, Ej, and Ek are the energies of primary radiation and
characteristic energies of the jth and kth elements, Ψ1, Ψ2
represent entrance and exit angles, G0 a geometrical
factor, I0 the intensity of primary radiation, ε the detector
efficiency at the fluorescent energy Ej, τ the photoelectric
mass absorption coefficient at primary energy, JK,L the
jump ratio for the appropriate (K or L) absorption edge, T
the transmission coefficient of the fluorescent X-rays in
the air path, f the fractional intensity of the X-ray line
used for analysis, e.g Kα/(Kα + Kβ). µ(E0), µ(Ej), µ(Ek) are
the total sample mass absorption coefficients for X-ray Figure 1. Schematic diagram of the experimental setup.
 Analysis of stainless steel samples by EDXRF spectrometry 635

concentrations were prepared and analysed and I0G was
found to be reproducible within ± 2%. All synthetic
standards were prepared as pellets of diameter 24⋅7 mm
from appropriate oxides. Quantitative analysis of all the
SS samples was performed by the computer program
CATXRF which is based on FP method and uses an
iterative procedure to calculate the concentration of
various elements present in the specimen. This program is
written in FORTRAN-77 and runs on Pentium PCs.
Details of the program are given elsewhere (Tiwari and
Sawhney 2000). We have detected almost all the elements
with Z > 17 present in steel samples, and the combined
concentration of all other elements (C, P, Si, S etc) which
cannot be detected by EDXRF spectrometry, is less than
1%. Hence no significant error is introduced in the
elemental analysis if the presence of these low-Z elements
is ignored. Accordingly it is assumed in our EDXRF
quantitative analysis that the total concentration of all
elements detected by EDXRF is 100%. In the iterative
procedure in CATXRF program, for the starting values,
the specimen was assumed to comprise of Fe and
cellulose in 1 : 1 ratio and the iteration was stopped when
total concentration of all detected elements became 100%.
The values of various fundamental parameters were taken
from published literature (McMaster et al 1969; Bam-
bynek et al 1972).
4. Results and discussion
Figure 2a shows a typical fluorescence spectrum of a T-
216 reference sample obtained with EDXRF spectro-

T-216 sample

Mn K α +Cr K β Fe K α Mo K α
)
3

Cr K α Source Scattered peaks

Ni K α Nb K α
Counts ( X10

Mo K β
(a)
Ni K β

5 10 15 20
Energy (KeV)

Mn K α Cr K β Mo K α
(b) (d)
10
3
3 Nb K α
Counts

10
Counts

2
10

2
10 1
10
8 8
(c) (e)
Residuals

4
Residuals

0 0

-4 -4

-8 -8

5.7 5.8 5.9 6.0 6.1 15.5 16.0 16.5 17.0 17.5 18.0

Energy (KeV) Energy (KeV)

Figure 2. (a) A typical EDXRF spectra of T-216 SS-reference sample, (b) merged Mn–Kα and Cr–Kβ X-ray peaks fitted using
NL-LSF procedures, (c) residuals of the fit in (b) in units of one standard deviation (σ = N , where N is the counts accumulated
in a given energy channel) of data; similarly (d) and (e) show fitting results obtained for Nb–Mo pair.
636 M K Tiwari, A K Singh and K J S Sawhney
Fe K α
4
10
Cr K α
Fe K β
3 Ni K α
10
Counts
Ni K β
2
10
5 10
Energy (KeV)
Figure 3. A typical EDXRF spectra of SS-304 sample.
meter. It is evident from the figure that the spectrum is
quite complex with several peaks strongly overlapping.
Owing to the complexity of the SS spectrum, the least
squares fitting was done by dividing the spectrum into
several regions. For instance figure 2b shows the Cr–Mn
part of the spectrum for which the individual peak areas
could only be derived by the use of NL-LSF procedure.
The goodness of the fit is evident from figure 2c where
the residuals of the fit are seen to fall within the ± 3σ
limit (Van Grieken and Markowicz 1993). Similarly the
fitting results obtained for the Nb–Mo part are shown in
figures 2d and e. The Nb–Mo peaks fall on the tail of the
broad Compton scattered source peak and hence the least
squares fitting is not that perfect in this region which is
reflected in relatively higher fitting residuals (figure 2e).
Barrea and Mainardi (1998) could not determine the
concentrations of minor elements like Mn, Co, Cu, Nb,
etc because of strong overlaps and also non-correction of
this type of overlaps (Kβ line of one element merging in
Kα intensity of another element) results in greater error in
their analysis. On the other hand we have been able to
determine accurately the concentrations of all elements
detectable by EDXRF technique.
The concentrations of minor and major elements in
various SS samples analysed by EDXRF technique are
given in tables 1 and 2. Tables show that the concentra-
tion of all elements in each sample matches quite well
SS-304 sample
Mo K α Source scattered peaks
Mo K β
Nb K α
15 20 25
with their certified values. The overall relative analysis
error range from 2% to 10%, which include errors
contributed by geometrical reproducibility, counting
statistics, errors in fundamental constants etc. The dominant
error is the statistical error in the peak intensity
determination which is higher for elements present in
lower concentration. Various SS samples analysed by
EDXRF technique were also analysed by VES (Vacuum
Emission Spark Spectrometer, Model JY56E, Jobin Yvon,
France). VES is an instrumental technique which is used
in the elemental analysis of SS types of samples. The
results of this analysis are also given in tables 1 and 2.
These tables show that the EDXRF results are similar to
VES results and in many cases the EDXRF results match
with certified values better than VES results. VES
technique is a destructive technique as it creates a
spark spot on the sample surface and also requires several
type-standards similar in concentration and form
of the unknown samples for the calibration of the
spectrometer. On the other hand, the method reported
here based on EDXRF technique does not require
any type-standards and is applicable for all types of
steel samples. The main advantage of VES technique
is that it also gives the concentrations of low Z elements
like C, P, Si etc which are not detectable by EDXRF
measurements. For most of the routine analysis of SS
the analysis can be performed by EDXRF, which is
 Analysis of stainless steel samples by EDXRF spectrometry 637

Table 1. Concentrations of CRM samples determined by EDXRF and VES.

Concentrations (%)
Sample
no. Elements Certified By EDXRF By VES

T-211 Cr 19⋅50 19⋅1 ± 0⋅40 18⋅4 ± 0⋅30

Mn 1⋅80 1⋅84 ± 0⋅08 1⋅60 ± 0⋅05
Fe 69⋅56 69⋅9 ± 1⋅60 69⋅9 ± 1⋅00
Ni 8⋅61 8⋅4 ± 0⋅20 8⋅6 ± 0⋅10
Cu 0⋅20 1381 ± 50 ppm –
Mo 0⋅20 0⋅21 ± 0⋅02 –
T-212 Cr 21⋅80 21⋅4 ± 0⋅50 21⋅4 ± 0⋅30
Mn 0⋅83 0⋅9 ± 0⋅05 0⋅77 ± 0⋅02
Fe 65⋅40 65⋅1 ± 1⋅50 67⋅6 ± 1⋅00
Co 0⋅18 1630 ± 150 ppm –
Ni 7⋅12 7⋅1 ± 0⋅20 6⋅5 ± 0⋅10
Cu 1⋅37 1⋅36 ± 0⋅05 –
Nb 0⋅01 192 ± 18 ppm –
Mo 2⋅50 2⋅5 ± 0⋅05 2⋅3 ± 0⋅05
T-213 Cr 21⋅60 22⋅0 ± 0⋅50 21⋅6 ± 0⋅33
Mn 1⋅07 1⋅0 ± 0⋅05 1⋅0 ± 0⋅03
Fe 64⋅90 65⋅2 ± 1⋅50 67⋅3 ± 1⋅00
Co 0⋅26 0⋅4 ± 0⋅04 –
Ni 7⋅24 7⋅2 ± 0⋅20 6⋅2 ± 0⋅10
Cu 1⋅50 1⋅5 ± 0⋅05 –
Nb 0⋅03 417 ± 40 ppm –
Mo 2⋅53 2⋅56 ± 0⋅06 2⋅37 ± 0⋅05
T-214 Cr 20⋅12 20⋅0 ± 0⋅40 20⋅3 ± 0⋅30
Mn 1⋅45 1⋅49 ± 0⋅06 1⋅43 ± 0⋅04
Fe 47⋅80 47⋅9 ± 1⋅10 48⋅6 ± 0⋅70
Co 0⋅29 0⋅37 ± 0⋅03 –
Ni 24⋅44 23⋅9 ± 0⋅60 23⋅9 ± 0⋅40
Cu 1⋅42 1⋅49 ± 0⋅05 –
Nb 0⋅01 226 ± 21 ppm –
Mo 4⋅24 4⋅28 ± 0⋅09 4⋅53 ± 0⋅10
T-215 Cr 18⋅63 19⋅4 ± 0⋅40 18⋅7 ± 0⋅28
Mn 1⋅38 1⋅4 ± 0⋅06 1⋅50 ± 0⋅05
Fe 65⋅40 65⋅0 ± 1⋅50 65⋅7 ± 1⋅00
Co 0⋅01 1451 ± 133. ppm –
Ni 10⋅95 10⋅9 ± 0⋅30 10⋅6 ± 0⋅20
Cu 0⋅02 495 ± 18⋅0 ppm –
Nb 0⋅76 0⋅81 ± 0⋅07 –
Mo 2⋅20 2⋅3 ± 0⋅05 2⋅16 ± 0⋅04
T-216 Cr 23⋅40 23⋅0 ± 0⋅53 23⋅6 ± 0⋅35
Mn 5⋅60 5⋅8 ± 0⋅25 5⋅6 ± 0⋅14
Fe 50⋅76 50⋅3 ± 1⋅10 50⋅9 ± 0⋅70
Co 0⋅16 0⋅25 ± 0⋅02 –
Ni 16⋅30 16⋅7 ± 0⋅40 15⋅2 ± 0⋅20
Cu 0⋅15 0⋅21 ± 0⋅01 –
Nb 0⋅24 0⋅24 ± 0⋅02 –
Mo 2⋅92 3⋅0 ± 0⋅10 2⋅9 ± 0⋅06
CRM-474 Cr 19⋅06 19⋅8 ± 0⋅45 18⋅7 ± 0⋅28
Mn 1⋅70 1⋅68 ± 0⋅07 1⋅71 ± 0⋅04
Fe 60⋅05 59⋅5 ± 1⋅40 60⋅5 ± 1⋅00
Ni 14⋅70 14⋅8 ± 0⋅40 15⋅1 ± 0⋅30
Cu 0⋅35 0⋅35 ± 0⋅01 –
Nb 0⋅01 157 ± 14⋅0 ppm –
Mo 3⋅55 3⋅7 ± 0⋅08 3⋅71 ± 0⋅07
CRM-475 Cr 14⋅14 14⋅4 ± 0⋅30 14⋅3 ± 0⋅22
Mn 0⋅89 0⋅86 ± 0⋅02 0⋅96 ± 0⋅02
Fe 75⋅00 75⋅3 ± 1⋅70 77⋅0 ± 1⋅00
Co 0⋅22 0⋅3 ± 0⋅03 –
Ni 5⋅66 5⋅7 ± 0⋅15 5⋅63 ± 0⋅10
Nb 0⋅22 0⋅23 ± 0⋅02 –
Mo 1⋅59 1⋅61 ± 0⋅04 1⋅6 ± 0⋅03
638 M K Tiwari, A K Singh and K J S Sawhney

Table 2. Concentrations of SS-304 and SS-316 samples determined by EDXRF and VES.

Concentrations (%)
Sample
no. Elements Certified range By EDXRF By VES

SS-304 Cr 18⋅0–20⋅0 18⋅9 ± 0⋅40 18⋅0 ± 0⋅27

Mn 2⋅0 Max. 1⋅23 ± 0⋅05 1⋅27 ± 0⋅03
Fe 60⋅0–70⋅0 69⋅9 ± 1⋅60 70⋅0 ± 1⋅00
Ni 8⋅0–12⋅0 8⋅9 ± 0⋅20 8⋅0 ± 0⋅10
Cu – 0⋅3 ± 0⋅01 –
Nb – 131 ± 20⋅0 ppm –
Mo – 0⋅6 ± 0⋅01 0⋅33 ± 0⋅01
SS-316 Cr 16⋅0–18⋅0 17⋅5 ± 0⋅40 16⋅8 ± 0⋅24
Mn 2⋅0 Max. 1⋅56 ± 0⋅07 1⋅34 ± 0⋅03
Fe 60⋅0–70⋅0 67⋅1 ± 1⋅52 67⋅4 ± 1⋅00
Co – 0⋅25 ± 0⋅02 –
Ni 10⋅0–14⋅0 10⋅9 ± 0⋅30 10⋅5 ± 0⋅20
Cu – 0⋅28 ± 0⋅01 –
Mo 2⋅0–3⋅0 2⋅3 ± 0⋅05 2⋅23 ± 0⋅04

seen to give good results by the method of analysis
reported in this paper.
In general, in XRF analysis of stainless steel types of
samples, where several high Z elements are present in
sufficiently high concentration, the enhancement effect
becomes quite dominant and can give erroneous results if
it is not corrected for mathematically. For example in SS,
Cr is strongly enhanced by Fe and can yield inflated
values for Cr concentration by as much as 40–60% if not
properly corrected. In our CATXRF program both matrix
absorption and enhancement effects are properly corrected
as per the equations given in the theory section.
5. Conclusions
In this paper a simple method for non-destructive and
standardless analysis of stainless steel samples is
presented. The agreement between the calculated concentra-
tions and certified concentration values of various
elements in different stainless steel samples have demon-
strated that the compositional analysis of stainless steel
samples can be made reliably with a simple radioisotope
excited EDXRF system. The CATXRF quantitative
analysis program in conjunction with the non-linear least
square fitting method gives a simple, fast, standardless,
nondestructive method of SS analysis. The system
described has provided our laboratory with an inexpensive
alternative to the more commonly used VES system for
SS analysis. The method reported here can also be used
for materials other than steel.
References
Barrea Raul A and Mainardi Raul T 1998 X-ray Spectrom. 27
111
Bambynek W, Crasemann B, Fink R W, Freund H U, Swift
C D, Price R E and Rao P V 1972 Rev. Mod. Phys. 44 716
Claisse F and Quintin M 1967 Can. Spectrosc. 12 129
Lachance G R and Traill R J 1966 Can. Spectrosc. 11 43
McMaster W H, Delgrande N K and Hubbell J H 1969 Compi-
lation of X-ray cross sections, UCRL-50174 (Livermore:
Lawrence Radiation Laboratory, Univ. California) p. 353
Rasberry S D and Heinrich K F J 1974 Anal. Chem. 46 81
Sawhney K J S and Lodha G S 1989 Computers and
Geosciences 15 1115
Sawhney K J S, Tiwari M K, Singh A K and Nandedkar R V
1998 Proc. 6th national seminar on X-ray spectroscopy and
allied areas (eds) S K Joshi et al (New Delhi: Narosa) p. 130
Sparks C J Jr 1975 Adv. X-ray Anal. 19 19
Tiwari M K and Sawhney K J S 2000 CATXRF: A quantitative
analysis program for EDXRF spectrometry, CAT technical
report
Van Dyck P M, Torok Sz B and Van Grieken R E 1986 Anal.
Chem. 58 1761
Van Grieken R E and Markowicz A A 1993 Handbook of X-ray
spectrometry (New York: Marcel Dekker) p. 181

View publication stats